Content
- The most popular content by clicks are forum, homepage, outside content, and quizzes.
EDA
library(tidyverse)
library(ggplot2)
library(patchwork)
feat <- read.csv("../../data/clean/features.csv")Introduction
This data describes student performance in online learning through The Open University, a university based in the United Kingdom. Courses are delivered in a structured fashion entirely online, and students are assigned small tutoring groups to support their learning.
The visualizations here describe the relationships between some characteristics of students and courses, as well as the scores achieved.
Scores are assigned in three different ways: TMA (tutor-marked), CMA (computer-marked), and “Exam” which is the final exam for a course. Different courses are called modules, and are labeled with letter codes here such as AAA or GGG in the column code_module. Courses are offered at different times of year as well, and this is shown in code_presentation.
Student interaction with the course material online is described by how many times they click on a specific piece of content, for example sum_clicksum_glossary sums the number of times the student clicked the glossary for the course in question.
Student characteristics offered include gender, disability status, highest formal education level achieved, age group (very broad), deprivation/poverty level (IMD Band), and region of UK where they reside.
Insights
- Question of interest: how much do the student’s characteristics impact the score? What’s the predictive value? Is this program providing equal opportunities/learning to students of different backgrounds?
Course Type
- BBB and FFF are the most popular courses
- Gender distribution across subjects is quite unequal
Scoring
- Computer marked exams are noticeably different from others - modules GGG and BBB seem to have the really different CMA traits
- The quantity of final exams seems smaller than expected - students withdrawing before course completion?
- Module BBB has strikingly high number of score zero as compared to others - also a course that has more female students
Students
- Most students are under age 35
- There are students from across the deprivation spectrum, with a slight skew towards the lower end
- Gender is relatively consistent across this spectrum, but missing deprivation data is substantially more male
- Most have some education prior to the course, usually A-levels
- Ireland is the least represented region
Features
Some features recoded in python script.
Academic
The id_assessment column probably has internal correlations that would lend to grouping.
ggplot(feat, aes(x=as.factor(code_module)))+
theme_bw()+
geom_histogram(stat="count")+
geom_text(stat='count', aes(label=..count..), nudge_y = 1500)
Semester
ggplot(feat, aes(x=as.factor(code_presentation)))+
theme_bw()+
geom_histogram(stat="count")+
geom_text(stat='count', aes(label=..count..), nudge_y = 1500)
Type of grading
ggplot(feat, aes(x=as.factor(assessment_type)))+
theme_bw()+
geom_histogram(stat="count")+
geom_text(stat='count', aes(label=..count..), nudge_y = 1500)
Prior Attempts
ggplot(feat, aes(x=num_of_prev_attempts))+
theme_bw()+
geom_histogram(stat="count")+
geom_text(stat='count', aes(label=..count..), nudge_y = 5000)
ggplot(feat, aes(x=log(num_of_prev_attempts), group=as.factor(code_module), fill = as.factor(code_module)))+
facet_grid(assessment_type~.)+
theme_bw()+
geom_density(alpha = .5)
Current Credits
ggplot(feat, aes(x=studied_credits))+
theme_bw()+
geom_density()+
geom_vline(xintercept=75)
ggplot(feat, aes(x=studied_credits, group=as.factor(code_module), fill = as.factor(code_module)))+
facet_grid(assessment_type~.)+
theme_bw()+
geom_vline(xintercept=75)+
geom_density(alpha = .5)
Dates
These indicate the days since the module was taught
ggplot(feat, aes(x=date))+
theme_bw()+
geom_density()
ggplot(feat, aes(x=date, group=as.factor(assessment_type), fill = as.factor(assessment_type)))+
theme_bw()+
geom_density(alpha = .5)
ggplot(feat, aes(x=date, group=as.factor(code_module), fill = as.factor(code_module)))+
facet_grid(assessment_type~.)+
theme_bw()+
geom_density(alpha = .5)
Clicks
skimr::skim(feat[, c("sum_clicksum_dataplus","sum_clicksum_dualpane","sum_clicksum_externalquiz"
,"sum_clicksum_folder","sum_clicksum_forumng","sum_clicksum_glossary","sum_clicksum_homepage","sum_clicksum_htmlactivity"
,"sum_clicksum_oucollaborate","sum_clicksum_oucontent","sum_clicksum_ouelluminate","sum_clicksum_ouwiki","sum_clicksum_page"
,"sum_clicksum_questionnaire","sum_clicksum_quiz","sum_clicksum_repeatactivity","sum_clicksum_resource","sum_clicksum_sharedsubpage"
,"sum_clicksum_subpage","sum_clicksum_url")])| Name | …[] |
| Number of rows | 190508 |
| Number of columns | 20 |
| _______________________ | |
| Column type frequency: | |
| numeric | 20 |
| ________________________ | |
| Group variables | None |
Variable type: numeric
| skim_variable | n_missing | complete_rate | mean | sd | p0 | p25 | p50 | p75 | p100 | hist |
|---|---|---|---|---|---|---|---|---|---|---|
| sum_clicksum_dataplus | 0 | 1 | 2.98 | 9.54 | 0 | 0 | 0 | 0 | 143 | ▇▁▁▁▁ |
| sum_clicksum_dualpane | 0 | 1 | 0.97 | 3.51 | 0 | 0 | 0 | 0 | 68 | ▇▁▁▁▁ |
| sum_clicksum_externalquiz | 0 | 1 | 3.18 | 10.79 | 0 | 0 | 0 | 0 | 340 | ▇▁▁▁▁ |
| sum_clicksum_folder | 0 | 1 | 0.33 | 1.12 | 0 | 0 | 0 | 0 | 13 | ▇▁▁▁▁ |
| sum_clicksum_forumng | 0 | 1 | 380.95 | 704.88 | 0 | 57 | 178 | 422 | 13154 | ▇▁▁▁▁ |
| sum_clicksum_glossary | 0 | 1 | 3.48 | 39.14 | 0 | 0 | 0 | 0 | 2952 | ▇▁▁▁▁ |
| sum_clicksum_homepage | 0 | 1 | 330.82 | 370.80 | 0 | 119 | 226 | 416 | 7277 | ▇▁▁▁▁ |
| sum_clicksum_htmlactivity | 0 | 1 | 0.44 | 1.76 | 0 | 0 | 0 | 0 | 33 | ▇▁▁▁▁ |
| sum_clicksum_oucollaborate | 0 | 1 | 4.39 | 12.17 | 0 | 0 | 0 | 3 | 316 | ▇▁▁▁▁ |
| sum_clicksum_oucontent | 0 | 1 | 554.06 | 811.90 | 0 | 37 | 205 | 761 | 9308 | ▇▁▁▁▁ |
| sum_clicksum_ouelluminate | 0 | 1 | 2.48 | 11.03 | 0 | 0 | 0 | 0 | 317 | ▇▁▁▁▁ |
| sum_clicksum_ouwiki | 0 | 1 | 33.69 | 82.39 | 0 | 0 | 0 | 32 | 2117 | ▇▁▁▁▁ |
| sum_clicksum_page | 0 | 1 | 3.33 | 7.24 | 0 | 0 | 0 | 4 | 334 | ▇▁▁▁▁ |
| sum_clicksum_questionnaire | 0 | 1 | 3.82 | 9.36 | 0 | 0 | 0 | 0 | 89 | ▇▁▁▁▁ |
| sum_clicksum_quiz | 0 | 1 | 376.21 | 565.80 | 0 | 30 | 119 | 615 | 13032 | ▇▁▁▁▁ |
| sum_clicksum_repeatactivity | 0 | 1 | 0.00 | 0.03 | 0 | 0 | 0 | 0 | 3 | ▇▁▁▁▁ |
| sum_clicksum_resource | 0 | 1 | 49.55 | 76.08 | 0 | 18 | 34 | 62 | 5147 | ▇▁▁▁▁ |
| sum_clicksum_sharedsubpage | 0 | 1 | 0.00 | 0.07 | 0 | 0 | 0 | 0 | 6 | ▇▁▁▁▁ |
| sum_clicksum_subpage | 0 | 1 | 173.68 | 194.10 | 0 | 36 | 108 | 252 | 4345 | ▇▁▁▁▁ |
| sum_clicksum_url | 0 | 1 | 26.42 | 43.15 | 0 | 4 | 14 | 35 | 2134 | ▇▁▁▁▁ |
Quick test of the high-clicking outliers
DT::datatable(feat[feat$sum_clicksum_quiz > 10000,])DT::datatable(feat[feat$sum_clicksum_forumng > 10000,])Student traits
ggplot(feat, aes(x=as.factor(age_band)))+
theme_bw()+
geom_histogram(stat="count")+
geom_text(stat='count', aes(label=..count..), nudge_y = 1500)
ggplot(feat, aes(x=as.factor(gender)))+
theme_bw()+
geom_histogram(stat="count")+
geom_text(stat='count', aes(label=..count..), nudge_y = 1500)
ggplot(feat, aes(x=as.factor(imd_band)))+
theme_bw()+
geom_histogram(stat="count")+
geom_text(stat='count', aes(label=..count..), nudge_y = 1500)
ggplot(feat, aes(x=as.factor(highest_education)))+
theme_bw()+
geom_histogram(stat="count")+
geom_text(stat='count', aes(label=..count..), nudge_y = 1500)
ggplot(feat, aes(x=as.factor(region)))+
theme_bw()+
theme(axis.text.x = element_text(angle = 45, hjust = 1))+
geom_histogram(stat="count")+
geom_text(stat='count', aes(label=..count..), nudge_y = 1500)
Gender x Poverty
feat$imd_band <- factor(feat$imd_band, levels = c("0-10%","10-20","20-30%","30-40%","40-50%","50-60%","60-70%","70-80%","80-90%","90-100%"))
ggplot(feat, aes(x=imd_band, group=(gender), fill = (gender)))+
theme_bw()+
geom_histogram(stat="count", alpha = .5)+
geom_text(stat='count', aes(label=..count..), position = position_stack(vjust = 0.5))
Course x Poverty
feat$imd_band <- factor(feat$imd_band, levels = c("0-10%","10-20","20-30%","30-40%","40-50%","50-60%","60-70%","70-80%","80-90%","90-100%"))
ggplot(feat, aes(x=code_module, group=(imd_band), fill = (imd_band)))+
theme_bw()+
geom_histogram(stat="count", alpha = .5)+
geom_text(stat='count', aes(label=..count..), position = position_stack(vjust = 0.5))
Test Type x Gender
feat$test_plus <- paste(feat$code_presentation, feat$assessment_type)
ggplot(feat, aes(x=test_plus, group=(gender), fill = (gender)))+
theme_bw()+
theme(axis.text.x = element_text(angle = 45, hjust = 1))+
geom_histogram(stat="count", alpha = .5)+
geom_text(stat='count', aes(label=..count..), position = position_stack(vjust = 0.5))
feat$test_plus <- paste(feat$code_module, feat$assessment_type)
ggplot(feat, aes(x=test_plus, group=(gender), fill = (gender)))+
theme_bw()+
theme(axis.text.x = element_text(angle = 45, hjust = 1))+
geom_histogram(stat="count", alpha = .5)+
geom_text(stat='count', aes(label=..count..), position = position_stack(vjust = 0.5))
Outcome
General
ggplot(feat, aes(x=score))+
theme_bw()+
geom_density()
ggplot(feat, aes(x=score, group=as.factor(assessment_type), fill = as.factor(assessment_type)))+
theme_bw()+
geom_density(alpha = .5)
ggplot(feat, aes(x=score, group=as.factor(code_module), fill = as.factor(code_module)))+
facet_grid(assessment_type~.)+
theme_bw()+
geom_density(alpha = .5)
ggplot(feat, aes(x=score, group=as.factor(assessment_type), fill = as.factor(assessment_type)))+
facet_grid(code_module~.)+
theme_bw()+
geom_density(alpha = .5)
Gender impact
ggplot(feat, aes(x=score, group=as.factor(gender), fill = as.factor(gender)))+
theme_bw()+
geom_density(alpha = .5)
ggplot(feat, aes(x=score, group=as.factor(code_module), fill = as.factor(code_module)))+
facet_grid(gender~.)+
theme_bw()+
geom_density(alpha = .5)
ggplot(feat, aes(x=score, group=as.factor(gender), fill = as.factor(gender)))+
facet_grid(code_module~.)+
theme_bw()+
geom_density(alpha = .5)
ggplot(feat, aes(x=score, group=as.factor(gender), fill = as.factor(gender)))+
facet_grid(code_module~., scales = "free_y")+
theme_bw()+
geom_histogram(alpha = .5)
Poverty impact
Clearly not randomly missing
feat$imd_band <- factor(feat$imd_band, levels = c("0-10%","10-20","20-30%","30-40%","40-50%","50-60%","60-70%","70-80%","80-90%","90-100%"))
ggplot(feat, aes(x=score, group=(imd_band), fill = (imd_band)))+
theme_bw()+
geom_density(alpha = .5)
ggplot(feat, aes(x=score, group=(code_module), fill = (code_module)))+
facet_grid(imd_band~.)+
theme_bw()+
geom_density(alpha = .5)
ggplot(feat, aes(x=score, group=(imd_band), fill = (imd_band)))+
facet_grid(code_module~.)+
theme_bw()+
geom_density(alpha = .5)
ggplot(feat, aes(x=score, group=(imd_band), fill = (imd_band)))+
facet_grid(code_module~., scales = "free_y")+
theme_bw()+
geom_histogram(alpha = .5)
Prior Attempts > 0
Retries score lower on tests - indicating that having to retake it displays poor command of material.
feat$prev_over_zero <- ifelse(feat$num_of_prev_attempts > 0, 1,0)
ggplot(feat, aes(x=score, group=as.factor(prev_over_zero), fill = as.factor(prev_over_zero)))+
facet_grid(assessment_type~.)+
theme_bw()+
geom_density(alpha = .5)
ggplot(feat, aes(x=score, group=as.factor(num_of_prev_attempts), fill = as.factor(num_of_prev_attempts)))+
facet_grid(num_of_prev_attempts~.)+
theme_bw()+
geom_density(alpha = .5)
Current Credits > 100
Taking more than 100 credits at a time is bad for overall performance.
feat$cred_over_100 <- ifelse(feat$studied_credits > 100, 1,0)
ggplot(feat, aes(x=score, group=as.factor(cred_over_100), fill = as.factor(cred_over_100)))+
facet_grid(assessment_type~.)+
theme_bw()+
geom_density(alpha = .5)
Test Type Detail
feat$test_plus <- paste(feat$code_presentation, feat$assessment_type)
ggplot(feat, aes(x=score, group=(test_plus), fill = (test_plus)))+
facet_grid(code_presentation~.)+
theme_bw()+
theme(axis.text.x = element_text(angle = 45, hjust = 1))+
geom_density(alpha = .5)
feat$test_plus <- paste(feat$code_module, feat$assessment_type)
ggplot(feat, aes(x=score, group=(test_plus), fill = (test_plus)))+
facet_grid(code_module~.)+
theme_bw()+
theme(axis.text.x = element_text(angle = 45, hjust = 1))+
geom_density(alpha = .5)
feat$test_plus <- paste(feat$code_module, feat$code_presentation)
ggplot(feat, aes(x=score, group=(test_plus), fill = (test_plus)))+
facet_grid(code_module~.)+
theme_bw()+
theme(axis.text.x = element_text(angle = 45, hjust = 1))+
geom_density(alpha = .5)